Eigen and Armadillo both are excellent linear algebra library. Armadillo provides not only mat operations but also similar functions with matlab, such as unique() and hist(). And Armadillo provides a syntax conversion table. It is very convenient. Dose Eigen provides functions and a detailed document similarly?

There is one.
Cite from http://eigen.tuxfamily.org/dox/ 3rd paragraph:
"For Matlab users, there is also a ASCII quick reference with Matlab translations."
-> http://eigen.tuxfamily.org/dox/AsciiQuickReference.txt

Anyone with up to date comparison?

Thank You.

One of the main difference is still up-to-date: Eigen comes with built-in and fast matrix products and linear matrix decompositions, whereas Armadillo needs to link to external BLAS/Lapack libraries for that. Another major difference is that Eigen is fully optimized for both small fixed-size matrices (like 3x3 or 4x4) and large dynamic dynamic sized ones, whereas Arma supports only the later.

Fantastic.

So Eigen is the right choice for me.

I appreciate some one handling small to medium matrices (Dynamic and Static).

Having using both I can say that - at least for sparse linear systems - Eigen performs much better than Armadillo.

Precisely through that conversion reference that was posted above, you can learn very fast how to port or start developing your application with Eigen.

Eigen also supports many types of Linear System Solvers like UMFPack (from MatLab) and SuperLu.

I have been using it quite a while and I am quite satisfied with the API syntax and speed.

Eigen is more elegant and efficient, I love Eigen. In my lab, we build many FEM programs based on Eigen, and these projects benefit so much from Eigen.

Great information guys.
Keep it coming.

I for one look for library which is the fastest for small - medium matrices (Up to ~1000 elements).
The problem is there is not enough data (Benchmarks) for those size on the net.

By the way, don't you think Eigen will get much more exposure and developers if it was on GitHub?